Diabetes Prediction App

Problem

How can machine-learning models help estimate diabetes likelihood from patient health data — and how do you make that accessible through a web interface?

Approach

Built a Django web application integrating ensemble ML methods. The system preprocesses patient data, runs predictions through trained models, and presents results through a clean web interface with report views.

Architecture

Python/Django backend with scikit-learn ensemble models. Data preprocessing pipeline handles missing values and normalization. Web interface built with Django templates for simplicity and reliability.

Technology

Chosen for reliability, maintainability and compatibility with standard hosting environments.

Key Features

• Ensemble ML model with 82% evaluation accuracy
• Real-time prediction through web interface
• Data preprocessing pipeline for health data
• Report generation and visualization

Challenges

Handling imbalanced medical datasets, selecting optimal ensemble parameters, and making ML predictions accessible to non-technical users through a clean interface.

Outcome

Final year project at BUBT — Top Position, BUBT Intra ML Competition (26 teams). Demonstrates end-to-end ML application development from data pipeline to production web interface.